Protist diversity—454 or Illumina？

[lingnanwu]

Hi,I'm a newbie in NGS.
Our groups are going to study the diversity of protist community,the target region—18s rRNA V9 region，length is about 150bp.I think Illumina is enough,but in many articles they all use 454.So I want to know if some one knows why they all use 454,which sequencing technology should I choice?Thank you.

[rglover]

The 454 error rate is much less than illumina - I'm assuming you'll want to assess the number of species etc.? There have been a few papers I think saying that illumina will over-estimate the number of species 10-100 fold because of the error rate of the short reads. We only ever use 454 for our amplicon studies.

[cliffbeall]

Illumina vs. 454

I would think Illumina might be a good option. You won't need the primary advantage of 454 which is the 400-500 bp read length. Cost per base will favor Illumina in a major way, and potentially you could get higher accuracy by sequencing the entire amplicon in both directions.

The possible drawback is that Illumina will require more method development -not as many people have used it for diversity studies (though I think that is changing).

[Jean]

Illumina error is LOWER than 454 as far as I remember. The problem is the PCR error that you will see pop up more in Illumina typically only because you are sequencing much deeper than 454.

I also think by now Illumina has been well-developed for 16s/18s studies. Consider doing overlapping paired reads to get the length you need.

[mcnelson.phd]

Hi lingnanwu,

Illumina would work perfectly for you intended application. You could use one of the published 16S protocols (Caporaso et al. and Bartram et al. are good places to start) and adapt it as necessary to your primers for the 18S. Given the size of your amplicon 454 would be a waste given the much greater throughput of Illumina. With paired-end reads, you could assemble them into a single contig quite easily regardless of which system you used (HiSeq currently does 2x100bp max reads and MiSeq currently does 2x150bp max)

The actual per-base sequencing error rates for Illumina are much lower than they are for 454. I've been trying to figure out where this "Illumina has significantly higher error rates than 454 for 16S" mantra started from, and I can't find any document evidence to support the claim. I suspect as Jean stated that it's a combination of people trying 16S on a HiSeq and the massive depth combined with PCR and minor sequencing errors lead to a much higher OTU count than expected. It's also quite possible that they didn't set up their runs properly by making sure the overall base composition was even by spiking in enough phiX. The fact that Rob Knight's group and, I've heard, the MBL are moving entirely to Illumina for their 16S needs means that there is no significant problem for Illumina compared to 454.